Research On Soft Sensor And Comprehensive Optimization Method In Grinding Control

Grinding process is a key and complicated link in ore dressing operation.The quality of its grinding particle size is closely related to the quality of the subsequent smelting metal and the recovery and utilization of metal.But traditional instrument inspection is difficult to satisfy the requirement of on-line grinding particle size inspection.In order to ensure high efficiency and stable size,only relying on the experience of worker can not satisfy the actual needs.At present,the optimal method of grinding soft measurement and grinding process optimization is the best way to solve this problem.Therefore,the paper has carried on the following three aspects in view of the above two methods.Firstly,the soft sensor model of grinding particle size based on neural network is established.The weight of neural network algorithm is optimized by crowd search algorithm.Considering the actual demand of grinding soft measurement,this paper improves the search step of the crowd search algorithm,and optimizes the calculation of membership function.At the same time,the Binomial Crossover Operator algorithm is used to update the location,which improves the population diversity.Through the above improvement,the search speed of the crowd search algorithm is improved.The simulation results show that the soft sensor algorithm can determine the neural network model more quickly,and meet the requirements of online soft measurement of grinding particle size.Secondly,the optimization problem is converted into parameter identification problem.This paper uses relaxation algorithm to identify the grinding comprehensive optimization model.The theory of relaxation is proved to be unbiased and effective.Since the relaxation algorithm is only suitable for off-line identification,in order to ensure the model parameters online correction and avoid the problem of data redundancy,a recursive relaxation algorithm is derived.The simulation results show that the proposed algorithm can identify the effective parameters online and accurately in the multi input multi output model.Finally,the algorithm of grinding soft sensor and the integrated optimization algorithm are verified on the semi physical simulation platform of grinding.Including data acquisition,grinding algorithm package,identification algorithm simulation,and comparison of the optimization results which verifies the effectiveness of the proposed method.